rohan anil
@_arohan_

co-founder, board @coreauto - and continuing to aspire to understand deep learning.

Joined December 2017
1,128 Photos and videos
Pinned Tweet
rohan anil
@_arohan_
Apr 19
It turns out multi step backpropaganda is better. paper has a beautiful way of improving backpropagation. One iteration cleanly gets us backprop, multiple iterations get us a preconditioned update.
rohan anil
@_arohan_
Apr 19
Replying to @LinYorker @ryu0000000001 @weijie444
arxiv.org/abs/2106.06199 Same update here
5
12
227
165,398
Nothing says" I've got my life together" like 1,000 lbs of salmon-fueled confidence.
Large brown bear standing on shoreline.

ALT Large brown bear standing on shoreline.

113
1,508
21,958
254,791
Muse when executed very well and scaled well could be bigger than AI mode or search.
7
88
7,567
A blast from the past @rtwlz found this website I made as a 14 year old! I am surprised by writing which hasn’t changed much like use of etc.
5
61
4,614
Found an indian bakery near my house. I need to resist.
40
2
279
23,825
I like Will, but my bet is on Riley. I trust him with my bank account.
JT
@jiratickets
Sep 24
the duality of man
5
148
21,185
Claude generated videos (particularly about p doom) was impressive the day it launched, now it feels like slop as more people are producing it.
5
2
69
4,730
rohan anil retweeted
josh
@eudaemonea
Sep 23
when Anthropic released their Functional Emotions paper, I gave it to Claude and asked for a song. tonight I asked Opus 5.5 to create a video for it. and it's breathtaking.
170
274
1,920
450,512
If Computational depth was the answer what was the question?
13
1
87
11,426
Shoot for the stars literally.
Riley Walz
@rtwlz
Sep 23
It's a little hard to internalize that the list of things one person can achieve is rapidly growing. You must think big to match what you're truly capable of now
5
76
7,849
rohan anil retweeted
Riley Walz
@rtwlz
Sep 23
It's a little hard to internalize that the list of things one person can achieve is rapidly growing. You must think big to match what you're truly capable of now
36
194
2,742
216,952
rohan anil retweeted
NotinReality@other__reality
Sep 22
Claude Opus 5.5 has the best visual design of any model I have tested so far
deckard
@slimer48484
Sep 9
Claude-Pop - I'm Upping My P(Doom)
264
675
6,470
2,313,003
Opus 5.5 feels like 4.5 when it was first released! Not much claudish and feels good. Very good model sir.
6
242
9,510
I got one shotted by a meme generator that I built. Was missing memegen
3
42
3,913
rohan anil retweeted
Jerry Tworek
@MillionInt
Sep 22
The road to singularity will be paved with blackberries of our generation. It is easier than ever to achieve a glimpse of growth and success but building a long term viable company is as hard as it ever was.
6
7
172
8,012
rohan anil retweeted
Thomas Wolf
@Thom_Wolf
Sep 21
releasing many high quality open-source RL environments is the most impactful thing anyone can do to push the open-source frontier right now the equivalent of sharing high quality pretraining data but in the new RLVR paradigm
elie
@eliebakouch
Sep 21
the most insane part, they will release ~7k RL training data and the framework leading to this top 6 model on AA, they also shipped the model + tech report less than 1 week after starting the final RL run pushing both intelligence and openness level, huge congrats
21
55
553
64,251
Memes can be automated. Removing gradient descent entirely is a bit harder
5
119
6,189
Good morning! ☀️
9
6
291
13,050
Curious if this has the same issue where being a good software engineer makes it perform worse due to brittle tests that Claude release recently pointed out.
wh
@nrehiew_
Sep 21
Looks like the MiMo RL runs have completed. The pro model went from 58.41 to 72.57 on DeepSWE. For context, the highest score on DeepSWE is 74 by Astra, Gemini 3.8 Flash and Opus 5
2
1
96
15,374
Load more